Better Climate Data Produces More Cost-Effective Pavement Designs

A recent study funded by the Mississippi DOT shows that the Mechanistic-Empirical Pavement Design Guide (MEPDG) is sensitive to climate inputs. "Better climate data makes a big difference in pavement performance predictions," says Mike Heitzman, assistant director at NCAT.

The MEPDG contains limited and incomplete climate data—only 12 sites for the entire state of Mississippi, with observations covering only 5 to 10 years. In order to adequately represent long-term statewide climate variations, more extensive climate data—both a denser site distribution and a broader time frame—were needed. NCAT partnered with Mississippi State University and Iowa State University's Climate Science Initiative to generate complete 40-year historic and future climate files for the state of Mississippi and evaluate these files' use in the MEPDG.

Building 40-year Historic Climate Files

The first step toward building 40-year historic climate files was collecting an extensive archive of available weather observations. The MEPDG requires hourly data for the following variables: air temperature, wind speed, percent sunshine, precipitation and relative humidity. For the period of interest (1970-2009), data were available from two sources:

  • Automated Surface Observation System (ASOS) and Automated Weather Observation System (AWOS)—These programs are maintained by government agencies and report a variety of hourly data; however, locations are limited primarily to airports, and the data is subject to errors and gaps in information. For this study, data from neighboring states' ASOS/AWOS sites was also gathered to aid in the analysis.
  • Cooperative Observer Program (COOP)—these observation sites, administered by the National Weather Service, are relatively dense compared to ASOS/AWOS. However, while the data is high quality, it is limited to daily high/low temperatures and precipitation totals.

Not every county within Mississippi has both an ASOS and a COOP site. Thus, a grid was constructed, with a grid spacing of approximately 25 km, so that there was roughly one grid point per county. The analysis procedure used natural neighbor interpolation, a spatial interpolation technique that weighs the relative contribution of surrounding observation sites on a particular point of interest. Separate analyses were performed for the hourly ASOS/AWOS data and the daily COOP data.

For each county, hourly values from the grid point closest to the centroid of the county were compiled. Temperature and precipitation values were also adjusted to include the higher-quality COOP data while maintaining the hourly variability determined using ASOS/AWOS data. An example is shown in Figure 1, in which an observed temperature curve (based on ASOS/AWOS data) is shifted to match the corresponding COOP high and low temperatures. In this manner, complete 40-year historic climate files were generated for each county within the state. These historic climate files are critical in matching climate characteristics with pavement data used in local MEPDG calibration.

daily downscaling example chart

Figure 1  Shift in temperature curve to reflect more accurate COOP data.

Building 40-year Future Climate Files

Well-constructed future climate scenarios aid in evaluating the impact of expected climate change on pavement performance. Global climate models, which reflect rising levels of greenhouse gases within the atmosphere, provide large-scale results (typically one grid point per 35,000 square miles). Dynamical downscaling is then used to refine the results spatially by means of a regional climate model. This study used the regional climate model RegCM2, with a grid-point spacing of approximately 80 miles.

Historic temperature and precipitation data is influenced by cyclic weather phenomena such as El Niño and La Niña, as well as major weather events such as hurricanes. However, the global climate model does not reflect these important influences on climate variability. This problem was addressed by first creating two proxy climate scenarios, using the global and regional models—a contemporary climate representing the 1990s and a future climate representing the 2040's. Both scenarios had realistic seasonal and daily temperature variations, but when compared with recorded historic data, the contemporary model reflected a systemic bias of ± 1-2°C. Assuming the bias was similar for both scenarios, subtracting the contemporary values from the future values (which were higher due to increasing greenhouse gases) resulted in eliminating or greatly reducing the bias. The difference in future and contemporary temperature values—and likewise, the ratio of future to contemporary precipitation—for each grid point and each month resulted in climate deltas. These climate deltas were then applied to the historic climate values. In this manner, the 40-year future climate file reflected both the predicted climate change due to increased greenhouse gases (climate deltas obtained from global and regional models), as well as the influence of cyclic and extreme weather events (from the historic values).

Impact on Pavement Performance Prediction

The effect of climate data on MEPDG pavement performance predictions was then evaluated. Several pavement designs were assessed using the 40-year historic climate scenario, the 40-year future climate scenario and the MEPDG climate data. Although differences were also observed in cracking and IRI predictions, the most dramatic differences were seen in rutting predictions. The following rut prediction examples are based on a thin (4.5 inch) hot-mix asphalt (HMA) design.

In Figure 2, the difference between predicted 20-year rutting using the MEPDG climate data and the 40-year historic climate data is shown for each climate zone in Mississippi. For each zone, the rutting difference is positive, indicating that predicted rutting is higher when using the MEPDG climate data. Thus, using the MEPDG climate data would result in overdesign of the pavement structure, either in terms of thickness or quality of materials.

Comparison of predicted 20-year rutting using MEPDG climate data and 40-year historic data

Figure 2  Comparison of predicted 20-year rutting using MEPDG climate data and 40-year historic data.

As shown in Figure 3, the differences between 20-year predicted rutting using the 40-year historic and future climate scenarios are almost negligible. The negative values indicate that the virtual (future) climate input, which reflected increased temperatures, resulted in slightly higher rutting predictions than the historic data. However, the magnitude of the differences was small, indicating that similar results are obtained whether using the historic climate data or the more labor-intensive future climate scenario.

Figure 4 shows the predicted rutting over a 40-year period for Mississippi's Zone 4 using all three climate data sets. While predicted rutting is similar for the historic and future climate scenarios, there is a substantial increase in predicted rutting using the MEPDG climate data. Better climate data, such as the 40-year historic climate file developed in this study, results in more accurate rutting predictions, which translates into a more cost-effective pavement design.

Comparison of predicted 20-year rutting using historic and virtual (future) climate scenarios.

Figure 3  Comparison of predicted 20-year rutting using historic and virtual (future) climate scenarios.

Predicted rutting for Zone 4 using MEPDG, historic and virtual (future) climate data sets.

Figure 4  Predicted rutting for Zone 4 using MEPDG, historic and virtual (future) climate data sets.

For more information on this study or if you are interested in pursuing a similar study for another state or region, please contact Mike Heitzman at 334.844.7309 or mah0016@auburn.edu.